Potential Determinants for Radiation-Induced Lymphopenia in Patients With Breast Cancer Using Interpretable Machine Learning Approach

Hao Yu, Fang Chen, Ka-On Lam, Li Yang, Yang Wang, Jian‐Yue Jin, Aya El Helali, Feng‐Ming KongFrontiers in Immunology 2022 · DOI

Abstract

Radiation-induced lymphopenia is known for its survival significance in patients with breast cancer treated with radiation therapy. This study aimed to evaluate the impact of radiotherapy on lymphocytes by applying machine learning strategies. We used Extreme Gradient Boosting (XGboost) to predict the event of lymphopenia (grade≥1) and conduced an independent validation. Then, we induced feature attribution analysis (Shapley additive explanation, SHAP) in explaining the XGboost models to explore the directional contribution of each feature to lymphopenia. Finally, we implemented the proof-of-concept clinical validation. The results showed that the XGboost models had rigorous generalization performances (accuracies 0.764 and ROC-AUC 0.841, respectively) in the independent cohort. The baseline lymphocyte counts are the most protective feature (SHAP = 5.226, direction of SHAP = -0.964). Baseline platelets and monocytes also played important protective roles. The usage of taxane only chemotherapy was less risk on lymphopenia than the combination of anthracycline and taxane. By the contribution analysis of dose, we identified that firstly lymphocytes were sensitive to a radiation dose less than 4Gy; secondly the irradiation volume was more important in promoting lymphopenia than the irradiation dose; thirdly the irradiation dose promoted the event of lymphopenia when the irradiation volume was fixed. Overall, our findings paved the way to clarifying the radiation dose volume effect. To avoid radiation-induced lymphopenia, irradiation volume should be kept to a minimum during the planning process, as long as the target coverage is not compromised.

Introduction

Radiation therapy is a cornerstone of breast cancer treatment, but it frequently induces lymphopenia—a significant depletion of lymphocytes. As these cells are critical for immune surveillance, their loss is associated with poorer survival outcomes. Despite its clinical importance, the precise relationship between radiation dose, irradiated volume, and patient-specific factors (such as baseline blood counts and chemotherapy history) has remained poorly understood. This study leverages interpretable machine learning to identify the key clinical and dosimetric drivers of radiation-induced lymphopenia (RIL).

Methods and Materials

The study utilized a robust machine learning pipeline to analyze data from two cohorts at the University of Hong Kong-Shenzhen Hospital: a testing cohort of 589 patients (2015–2019) and an independent validation cohort of 203 patients (2019–2020).

Figure 1: The study flowchart illustrating (A) patient selection, (B) the machine learning pipeline, and (C) the statistical verification process.

The researchers employed XGBoost (Extreme Gradient Boosting) to predict lymphopenia (Grade ≥1), chosen for its ability to model complex, non-linear relationships. To ensure transparency, SHAP (SHapley Additive exPlanation) values were used to quantify the importance and directional influence of each feature. A proof-of-concept clinical validation using paired t-tests on matched patients was also conducted to confirm the model's findings.

Key Insights

Volume Over Dose

A primary finding of this research is that the irradiation volume is a more significant driver of lymphopenia than the radiation dose itself. Spreading a lower dose over a larger anatomical area appears more detrimental to the circulating immune system than a higher dose concentrated in a smaller field.

Extreme Radiosensitivity

The analysis revealed that lymphocytes are sensitive to radiation doses as low as 4Gy. This suggests that even "scatter" radiation or low-dose "baths" during treatment can lead to significant immune cell depletion.

Protective Factors and Chemotherapy Synergy

Baseline counts of lymphocytes, platelets, and monocytes served as critical "buffers" against RIL. Furthermore, the study noted a synergistic effect with chemotherapy: patients treated with a combination of anthracycline and taxane faced a higher risk of RIL compared to those receiving taxane alone.

Results

The XGBoost model demonstrated excellent generalization, achieving an accuracy of 0.764 and an ROC-AUC of 0.841 on the independent validation cohort. Feature importance analysis identified the baseline lymphocyte count as the most significant protective factor (SHAP = 5.226). Dosimetrically, the "integral dose of the total body"—a proxy for total energy absorbed—was the second most important feature, underscoring the systemic impact of radiation.

Conclusion

This study highlights the importance of minimizing irradiation volume to preserve the immune system in breast cancer patients. Future radiotherapy planning should prioritize reducing the "low-dose bath" to protect the circulating lymphocyte pool. While limited by its single-center design, this work provides a clear path toward more immune-sparing radiation protocols.

Summary updated

Abstract

Radiation-induced lymphopenia is known for its survival significance in patients with breast cancer treated with radiation therapy. This study aimed to evaluate the impact of radiotherapy on lymphocytes by applying machine learning strategies. We used Extreme Gradient Boosting (XGboost) to predict the event of lymphopenia (grade≥1) and conduced an independent validation. Then, we induced feature attribution analysis (Shapley additive explanation, SHAP) in explaining the XGboost models to explore the directional contribution of each feature to lymphopenia. Finally, we implemented the proof-of-concept clinical validation. The results showed that the XGboost models had rigorous generalization performances (accuracies 0.764 and ROC-AUC 0.841, respectively) in the independent cohort. The baseline lymphocyte counts are the most protective feature (SHAP = 5.226, direction of SHAP = -0.964). Baseline platelets and monocytes also played important protective roles. The usage of taxane only chemotherapy was less risk on lymphopenia than the combination of anthracycline and taxane. By the contribution analysis of dose, we identified that firstly lymphocytes were sensitive to a radiation dose less than 4Gy; secondly the irradiation volume was more important in promoting lymphopenia than the irradiation dose; thirdly the irradiation dose promoted the event of lymphopenia when the irradiation volume was fixed. Overall, our findings paved the way to clarifying the radiation dose volume effect. To avoid radiation-induced lymphopenia, irradiation volume should be kept to a minimum during the planning process, as long as the target coverage is not compromised.
BibTeX
@inproceedings{yu2022potential,
  author = {Hao Yu and Fang Chen and Ka-On Lam and Li Yang and Yang Wang and Jian‐Yue Jin and Aya El Helali and Feng‐Ming Kong},
  title = {{Potential Determinants for Radiation-Induced Lymphopenia in Patients With Breast Cancer Using Interpretable Machine Learning Approach}},
  booktitle = {Frontiers in Immunology},
  year = {2022},
  doi = {10.3389/fimmu.2022.768811},
}
Figure